Diophantine Inference on a Statistical Database

Abstract

A statistical database is said to be compromisable if individual data items can be inferred from queryable values of statistical aggregates (mean, maximum, count, etc.) ((Denning82), ch. 6). We discuss here some methods, which while only leading to compromise of individual records on occasion, do lead to powerful inferences of other statistical characteristics of a database which may also be sensitive information. These methods use a new technique that has not apparently heretofore been explored, solution of simultaneous Diophantine (integer-solution) equations.